Marketing Research, MKTG 756

1
Market SegmentationThe why, how, and what value
of identifying homogeneous groups of customers

• Marketing Research Lecture 20

2
The most fundamental Marketing concept STP

• A basic belief in Marketing is to deliver the
  right products, to the right people, in the right
  way.
• If that is done successfully, one will be able to
  attract a greater price premium and one will be
  able to retain customers better.

T Targeting
S Segmentation
P Positioning
3
What is a Market Segment?
Market segmentation is the subdividing of a
market into distinct subsets, where any subset
may conceivably be selected as a marketing
target to be reached within a distinct marketing
mix. Kotler
4
The Essence of Marketing is Discrimination
Propositions People like to have things that
exactly meet their needs, as opposed to things
that function OK, but do not precisely meet their
needs People like to be talked to in ways that
reflect full understanding of their needs, rather
than in more general terms People will often pay
more for things that exactly meet their needs,
and/or as a result of being talked to in
specific ways
5
Why Segment the Market? Variations in Marketing
Actions Based on Segment Differences
Products Product design and positioning more in
line with the needs of consumers in each segment.
Customer Communications Better directed media
selection and advertising messages Salesforce
allocation to key customers Customized sales
presentations
6
Distribution Selection of distribution channels
consistent with segment buying patters
Pricing More precise guidelines for price
discrimination by levels of price sensitivity in
each segment
7
Advantages of Segmentation
Specialist Cost Price Monopoly
8
However, it is infeasible to target each
individual customer.
It is hard to name a company that has made money
due to true one-to-one marketing.
Cost/Unit
Volume
9
Should All Markets Be Segmented?
Not if...
Homogenous Market Market is too Small Benefit lt
Cost New Product Monopoly Commodity (?)
Can you name some markets that should not be
segmented?
10
Basic Segmentation Issues
How do I segment the market, that is, how do I
break up the market into distinct
subsets? Todays Lecture How do I select
which segments to serve? Briefly here but next
Lecture Once decided, this will allow me to
EXPAND my customer base, DEFEND my customer base,
and cross-sell to my customer base.
11
Implementing Segmentation Steps
Market Segmentation 1. Identify bases for
segmenting the market 2. Develop profiles or
segments 3. Relate to descriptors for
reachability
IMPLEMENT
Target Marketing 4. Develop measures of segment
attractiveness 5. Select the target segments
Product Positioning 6. Develop product
positioning for each target segment 7. Develop
marketing mix for each target segment
12
Major Segmentation Variables
Demographics Geographic Purchasing
Approaches Personal Characteristics
(Psychographics)
Source Adapted from Thomas v. Bonama and Benson
P. Shapiro, Segmenting the Industrial Market
(Lexington, MA Lexington Books, 1983)
13
Favorite categories by genderDemographic
Segmentation
To target men v. women, which sites should I
advertise on? Might make sense if behavioral
differences is your basis.
S e g m e n t a t i o n
14
Geographic Which areas of the country have
lifestyles and needs most served by my product?
Distribution Play/Weather/Regional Differences
S e g m e n t a t i o n
15
Psychographics

• Definition the use of psychological,
  sociological, and anthropological factors
  combined with demographic information to identify
  market segments with a propensity to favor some
  product groups or characteristics over another
  due to the unique combination of these factors.

DEMOGRAPHICS PSYCHOGRAPHIC PROFILES
This is the basis for segmentation that
conceptually marketers agree upon.
16
How to Select Segments to Target

• Market Attraction
• Segment size
• Amount and strength of growth
• Value of segment
• Company Fit
• Current position in segment
• Ease of entry into segment
• Competitive Environment
• Number and strength of competitors
• Ease of competitive entry into segment

17
An intro to cluster analysis

• Cluster analysis seeks to group objects such
  that segments are created that are as homogenous
  as possible given the choices by the researcher
• Cluster analysis works on the principle of
  maximizing the between-cluster variance while
  minimizing the within cluster variance
• Every object is allocated to one cluster

18
Illustration in one dimension
Between group variance
Within group variance
19
Illustration in two dimensions
20
Basic idea behind clustering
Which variables?
How do I compute the distance between any two
persons?
objects
Assignment of objects to groups
21
Summary of decisions to make

• Which clustering variables are going to be
  included?
• Which distance metric?
• Which cluster method? (we will discuss this next)
• How many clusters?
• How do I name the cluster (the most important
  step)

22
Which clustering variables to include

• Typically people use ALL psychographic variables
  initially (big as an elephant)
• People do not typically use outcome measures
  (like sales, overall satisfaction, etc) to
  determine ones cluster.
• After fitting the cluster solution, you can see
  which variables are different between the
  clusters.
• Something we will talk about in a few weeks,
  discriminant analysis, will inform us as to which
  Xs are the useful separators, i.e. determine
  which cluster someone is in.

23
Cluster Analysis Basic Idea

• In order to group objects, some kind of
  similarity measure is needed.
• Similar objects are grouped together and those
  further apart are put in separate clusters.
• Euclidean distance
• Manhattan or Metropolis Distance
• Euclidean distance (appropriate for
  metric/continuous variables)
• Metropolis Distance (appropriate for nominal
  variables)

24
Which Cluster Method?

• Two most common clustering approaches
• Hierarchical clustering (tree-based clustering)
• Non-hierarchical clustering

25
Hierarchical Clustering Single (Average) Linkage

• The Single Linkage method is based on the
  shortest (average) distance. It finds the two
  objects separated by the shortest (average)
  distance and places them in the first cluster.
• Then the next shortest (average) distance is
  found, and either a third individual joins the
  first 2 to form a cluster or a new two-individual
  cluster is formed.
• The process continues until all individuals are
  in one cluster.
• Also called the nearest-neighbor approach.

26
K-means clustering

• K-means clustering is the most commonly used
  clustering technique.
• It is an iterative technique that seeks to
  allocate each observation to the cluster that is
  located closest to it
• The number of clusters is chosen by you, and you
  decide upon the correct number of clusters using
  either business criterion, or by statistical
  criterion using a scree plot.

Within Cluster Sum of Squares
of clusters
27
How many clusters?

• There are no formal significance tests for the
  number of clusters.
• There are informal tests that involve the amount
  of heterogeneity within each cluster (Scree plot)
• People typically decide based on managerial
  action
• (a) how many clusters are feasible to target?
• (b) are each of the clusters of sufficient size
  to be considered a market segment?
• (c) Is there sufficient differentiation between
  the segments?

28
Graphical description of K-means clustering with
2 variables
Price sensitivity
Quality
S e g m e n t a t i o n
29
Iteration 1

• Choose the centers of the three clusters randomly
• I made a rather bad choice on purpose
• Allocate each point to its nearest center

S e g m e n t a t i o n
30
Iteration 2

• Choose the centers again as the centroids of the
  clusters from iteration 1.
• Allocate each point to the centroid that is
  closest to it

S e g m e n t a t i o n
31
Iteration 3

• Choose the centers again as the centroids of the
  clusters from iteration 2.
• Allocate each point to the centroid that is
  closest to it

S e g m e n t a t i o n
32
Iteration 4 and final step

• Choose the centers again as the centroids of the
  clusters from iteration 3.
• Nothing changed!!
• Voila, youre done

S e g m e n t a t i o n
33
An Example DuPont Data

• DuPont Data
• 102 Customers
• 20 Questions
• http//mktgweb.wharton.upenn.edu/ebradlow/Mktg209
  66/dupont.doc

34
Results of Average Linkage Approach
The higher up two groups are joined, the
farther apart they are
35
Results of K-means clustering with K2

• K-Means Clustering
• Centers
• Q1 Q2 Q3 Q4 TS1
  TS2 TS3
• 1, 8.590361 8.795181 8.397590 8.710843 8.469880
  8.554217 8.566265
• 2, 7.789474 8.000000 7.263158 7.789474 5.947368
  6.157895 6.421053
• SM1 SM2 SD1 SD2 SD3
  SD4 SD5
• 1, 8.807229 8.783133 8.891566 8.939759 8.746988
  8.73494 8.662651
• 2, 5.947368 4.736842 8.263158 8.789474 6.210526
  6.00000 6.947368
• SD6 SD7 INN1 INN2 INN3
  
• 1, 9.108434 9.144578 8.072289 8.253012 8.301205
• 2, 8.578947 8.631579 4.315789 4.578947 6.526316
• Clustering vector
• 1 2 1 2 2 1 1 1 1 1 1 1 2 1 2 1 1 1 1 1 1 1 1
  1 2 1 1 1 1 1 1 1 1 1
• 34 1 2 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 2 1 2 1 1
  2 1 2 1 1 1 1 1 2 2 1
• 67 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 2
  1 1 1 1 1 1 1 1 1 1 2
• 100 1 2 1
• Within cluster sum of squares

How different are the variables between the
clusters?
Who is in which cluster?
How much error within the clusters?
How big are the clusters?
36
Profiling

• Once you determine the market segments, to make
  things actionable you want to compute who is in
  what segment (profiling)
• Here is where you use demographics and behaviors
  for each segment.
• Useful for reach and targeting purposes.
• Name the segment (ex Yuppie Young Urban
  Professional)

37
Useful Websites

• http//www.statsoftinc.com/textbook/stcluan.html
• http//obelia.jde.aca.mmu.ac.uk/multivar/ca.htm
• http//trochim.human.cornell.edu/tutorial/flynn/cl
  uster.htm
• http//www.dssresearch.com/library/segment/underst
  anding.asp

38
Summary of Market Segmentation

• Cluster people based on their needs
  (Psychographics, Part-worths, etc)
• Then use demographics and psychographics for
  reach
• The question then becomes what is the value of
  each segment. This is what we will talk about
  next time.